说明：  通过核 K- 均值聚类的方法对语音帧进行聚类 , 由于聚类的中心能够很好地代表类内的特征, 用中心样本帧取代该类, 减少了核矩阵的维数, 然后再采用稀疏 KPCA方法对核矩阵进行特征提取。
(Through the nuclear K-means clustering method for clustering of speech frames, the cluster center can be a good representative of the class characteristics of the sample frame to replace the class with the center, reducing the dimension of the nuclear matrix, and then use Sparse KPCA method for feature extraction of the nuclear matrix.)